Why a Chinese AI Lab’s Product Lead Thinks AGI is “Inevitable”
ChinaTalk Irene Zhang
Opinion — commentary, not a factual news event.
A Chinese AI product lead says AGI is inevitable, and the Hugging Face hack proves the field is still in discovery mode. His bigger point: open and closed models are already having to work together.
Based on reporting by ChinaTalk, Irene Zhang — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Richard Bian runs product operations for Ant Ling, the foundation model effort at Ant, the company best known for Alipay. He’s not sounding like a casual fanboy. He calls the current AI push a discovery process, compares it to the Age of Discovery, and says the Hugging Face incident involving OpenAI research agents was less a scandal than a preview of problems the industry was bound to hit anyway.
That incident matters, in his view, because it shows the limits of closed systems and the usefulness of open ones. Hugging Face eventually used an open model from Z.ai locally to process logs after frontier models hit cybersecurity guardrails. For Bian, that’s not a footnote. It’s proof that open-weight and closed-weight models will end up complementing each other, not replacing each other in a neat winner-takes-all story.
He’s also blunt about where this goes next. AGI is coming, he says, though nobody knows what form it will take, and the industry is still at the stage where it doesn’t even know what the product should be. Coding is useful, but he doesn’t see that as the final shape. He’s watching coding agents drift toward co-work agents and says the market is still basically at home plate.
Bian’s framework is less about benchmarks than about cycles. He thinks technology runs through roughly 10- to 12-year hype waves, with capital piling in and making the next phase self-fulfilling. AI, in his telling, has reset the game back to the start. Generative AI commoditizes knowledge the way the internet commoditized information, but the business question is now gross margins, not capex.
That’s why he keeps coming back to evaluation and ecosystem diversity. He thinks people lean too hard on leaderboards and too little on better ways to judge models. And he sees open source as a product strategy, not a charity project: a way to pull in global talent, move faster, and support more products at once. Ant’s latest move fits that story too — Ling 3.1 Flash was open-sourced on October 14, with 560B parameters and up to a 1M-token context window.
My take — AI-written commentary, not fact-checked reporting
This is the rare AI take that sounds less like a keynote and more like a lab notebook. The industry loves acting like every new model is a clean victory lap, but the ugly truth is that the open and closed camps are already forced to share the same road. That’s not a bug; it’s the whole story, and the people pretending otherwise are mostly selling something.
Read more about this at: ChinaTalk
